fix style and wrong point order (#2052)

Author: felixdittrich92

Makefile

@@ -13,6 +13,7 @@ style:
 test:
 	coverage run -m pytest tests/common/ -rs
 	coverage run -m pytest tests/pytorch/ -rs
+	coverage report --fail-under=80 --show-missing
 
 test-common:
 	coverage run -m pytest tests/common/ -rs

README.md

@@ -24,7 +24,7 @@ As such, you can select the architecture used for [text detection](https://minde
 ```python
 from doctr.models import ocr_predictor
 
-model = ocr_predictor(det_arch='db_resnet50', reco_arch='crnn_vgg16_bn', pretrained=True)
+model = ocr_predictor(det_arch="db_resnet50", reco_arch="crnn_vgg16_bn", pretrained=True)
 ```
 
 ### Reading files
@@ -33,6 +33,7 @@ Documents can be interpreted from PDF or images:
 
 ```python
 from doctr.io import DocumentFile
+
 # PDF
 pdf_doc = DocumentFile.from_pdf("path/to/your/doc.pdf")
 # Image
@@ -87,7 +88,9 @@ Or even rebuild the original document from its predictions:
 import matplotlib.pyplot as plt
 
 synthetic_pages = result.synthesize()
-plt.imshow(synthetic_pages[0]); plt.axis('off'); plt.show()
+plt.imshow(synthetic_pages[0])
+plt.axis("off")
+plt.show()
 ```
 
 ![Synthesis sample](https://github.com/mindee/doctr/raw/main/docs/images/synthesized_sample.png)
@@ -112,7 +115,7 @@ from doctr.io import DocumentFile
 from doctr.models import kie_predictor
 
 # Model
-model = kie_predictor(det_arch='db_resnet50', reco_arch='crnn_vgg16_bn', pretrained=True)
+model = kie_predictor(det_arch="db_resnet50", reco_arch="crnn_vgg16_bn", pretrained=True)
 # PDF
 doc = DocumentFile.from_pdf("path/to/your/doc.pdf")
 # Analyze
@@ -313,7 +316,7 @@ import requests
 
 params = {"det_arch": "db_resnet50", "reco_arch": "crnn_vgg16_bn"}
 
-with open('/path/to/your/doc.jpg', 'rb') as f:
+with open("/path/to/your/doc.jpg", "rb") as f:
     files = [  # application/pdf, image/jpeg, image/png supported
         ("files", ("doc.jpg", f.read(), "image/jpeg")),
     ]

api/README.md

@@ -39,7 +39,7 @@ import requests
 headers = {"accept": "application/json"}
 params = {"det_arch": "db_resnet50"}
 
-with open('/path/to/your/img.jpg', 'rb') as f:
+with open("/path/to/your/img.jpg", "rb") as f:
     files = [  # application/pdf, image/jpeg, image/png supported
         ("files", ("117319856-fc35bf00-ae8b-11eb-9b51-ca5aba673466.jpg", f.read(), "image/jpeg")),
     ]
@@ -83,7 +83,7 @@ import requests
 headers = {"accept": "application/json"}
 params = {"reco_arch": "crnn_vgg16_bn"}
 
-with open('/path/to/your/img.jpg', 'rb') as f:
+with open("/path/to/your/img.jpg", "rb") as f:
     files = [  # application/pdf, image/jpeg, image/png supported
         ("files", ("117133599-c073fa00-ada4-11eb-831b-412de4d28341.jpeg", f.read(), "image/jpeg")),
     ]
@@ -115,7 +115,7 @@ import requests
 headers = {"accept": "application/json"}
 params = {"det_arch": "db_resnet50", "reco_arch": "crnn_vgg16_bn"}
 
-with open('/path/to/your/img.jpg', 'rb') as f:
+with open("/path/to/your/img.jpg", "rb") as f:
     files = [  # application/pdf, image/jpeg, image/png supported
         ("files", ("117319856-fc35bf00-ae8b-11eb-9b51-ca5aba673466.jpg", f.read(), "image/jpeg")),
     ]

api/tests/conftest.py

@@ -45,24 +45,24 @@ def mock_detection_response():
             "name": "117319856-fc35bf00-ae8b-11eb-9b51-ca5aba673466.jpg",
             "geometries": [
                 [
-                    0.8203927977629988,
-                    0.2021484375,
                     0.8203927977629988,
                     0.181640625,
                     0.906015010958283,
                     0.181640625,
                     0.906015010958283,
                     0.2021484375,
+                    0.8203927977629988,
+                    0.2021484375,
                 ],
                 [
-                    0.7467381290758103,
-                    0.20086261630058289,
                     0.7482568619833604,
                     0.17938309907913208,
                     0.8208542842026056,
                     0.1819499135017395,
                     0.8193355512950555,
                     0.2034294307231903,
+                    0.7467381290758103,
+                    0.20086261630058289,
                 ],
             ],
         },
@@ -208,26 +208,26 @@ def mock_ocr_response():
                     "blocks": [
                         {
                             "geometry": [
-                                0.9052250981330872,
-                                0.20474359393119812,
                                 0.7460642457008362,
                                 0.2017778754234314,
                                 0.7464945912361145,
                                 0.17868199944496155,
                                 0.9056554436683655,
                                 0.18164771795272827,
+                                0.9052250981330872,
+                                0.20474359393119812,
                             ],
                             "lines": [
                                 {
                                     "geometry": [
-                                        0.9052250981330872,
-                                        0.20474359393119812,
                                         0.7460642457008362,
                                         0.2017778754234314,
                                         0.7464945912361145,
                                         0.17868199944496155,
                                         0.9056554436683655,
                                         0.18164771795272827,
+                                        0.9052250981330872,
+                                        0.20474359393119812,
                                     ],
                                     "objectness_score": 0.54,
                                     "words": [

doctr/datasets/generator/base.py

@@ -66,7 +66,7 @@ def __init__(
     ) -> None:
         self.vocab = vocab
         self._num_samples = num_samples
-        self.font_family = font_family if isinstance(font_family, list) else [font_family]  # type: ignore[list-item]
+        self.font_family = font_family if isinstance(font_family, list) else [font_family]
         # Validate fonts
         if isinstance(font_family, list):
             for font in self.font_family:
@@ -116,7 +116,7 @@ def __init__(
         self.vocab = vocab
         self.wordlen_range = (min_chars, max_chars)
         self._num_samples = num_samples
-        self.font_family = font_family if isinstance(font_family, list) else [font_family]  # type: ignore[list-item]
+        self.font_family = font_family if isinstance(font_family, list) else [font_family]
         # Validate fonts
         if isinstance(font_family, list):
             for font in self.font_family:

doctr/models/builder.py

@@ -317,10 +317,8 @@ def __call__(
         ) != len(crop_orientations) != len(objectness_scores):
             raise ValueError("All arguments are expected to be lists of the same size")
 
-        _orientations = (
-            orientations if isinstance(orientations, list) else [None] * len(boxes)  # type: ignore[list-item]
-        )
-        _languages = languages if isinstance(languages, list) else [None] * len(boxes)  # type: ignore[list-item]
+        _orientations = orientations if isinstance(orientations, list) else [None] * len(boxes)
+        _languages = languages if isinstance(languages, list) else [None] * len(boxes)
         if self.export_as_straight_boxes and len(boxes) > 0:
             # If boxes are already straight OK, else fit a bounding rect
             if boxes[0].ndim == 3:
@@ -402,10 +400,8 @@ def __call__(  # type: ignore[override]
             page_shapes
         ) != len(crop_orientations) != len(objectness_scores):
             raise ValueError("All arguments are expected to be lists of the same size")
-        _orientations = (
-            orientations if isinstance(orientations, list) else [None] * len(boxes)  # type: ignore[list-item]
-        )
-        _languages = languages if isinstance(languages, list) else [None] * len(boxes)  # type: ignore[list-item]
+        _orientations = orientations if isinstance(orientations, list) else [None] * len(boxes)
+        _languages = languages if isinstance(languages, list) else [None] * len(boxes)
         if self.export_as_straight_boxes and len(boxes) > 0:
             # If boxes are already straight OK, else fit a bounding rect
             if next(iter(boxes[0].values())).ndim == 3:

doctr/models/detection/differentiable_binarization/base.py

@@ -11,6 +11,8 @@
 import pyclipper
 from shapely.geometry import Polygon
 
+from doctr.utils import order_points
+
 from ..core import DetectionPostProcessor
 
 __all__ = ["DBPostProcessor"]
@@ -80,7 +82,7 @@ def polygon_to_box(
         return (
             cv2.boundingRect(expanded_points)  # type: ignore[return-value]
             if self.assume_straight_pages
-            else np.roll(cv2.boxPoints(cv2.minAreaRect(expanded_points)), -1, axis=0)
+            else order_points(cv2.boxPoints(cv2.minAreaRect(expanded_points)))
         )
 
     def bitmap_to_boxes(

doctr/models/detection/fast/base.py

@@ -12,6 +12,7 @@
 from shapely.geometry import Polygon
 
 from doctr.models.core import BaseModel
+from doctr.utils import order_points
 
 from ..core import DetectionPostProcessor
 
@@ -78,7 +79,7 @@ def polygon_to_box(
         return (
             cv2.boundingRect(expanded_points)  # type: ignore[return-value]
             if self.assume_straight_pages
-            else np.roll(cv2.boxPoints(cv2.minAreaRect(expanded_points)), -1, axis=0)
+            else order_points(cv2.boxPoints(cv2.minAreaRect(expanded_points)))
         )
 
     def bitmap_to_boxes(

doctr/models/detection/linknet/base.py

@@ -12,6 +12,7 @@
 from shapely.geometry import Polygon
 
 from doctr.models.core import BaseModel
+from doctr.utils import order_points
 
 from ..core import DetectionPostProcessor
 
@@ -78,7 +79,7 @@ def polygon_to_box(
         return (
             cv2.boundingRect(expanded_points)  # type: ignore[return-value]
             if self.assume_straight_pages
-            else np.roll(cv2.boxPoints(cv2.minAreaRect(expanded_points)), -1, axis=0)
+            else order_points(cv2.boxPoints(cv2.minAreaRect(expanded_points)))
         )
 
     def bitmap_to_boxes(

doctr/models/detection/zoo.py

@@ -50,8 +50,8 @@ def _predictor(arch: Any, pretrained: bool, assume_straight_pages: bool = True,
             raise ValueError(f"unknown architecture: {type(arch)}")
 
         _model = arch
-        _model.assume_straight_pages = assume_straight_pages
-        _model.postprocessor.assume_straight_pages = assume_straight_pages
+        _model.assume_straight_pages = assume_straight_pages  # type: ignore[attr-defined]
+        _model.postprocessor.assume_straight_pages = assume_straight_pages  # type: ignore[attr-defined]
 
     kwargs.pop("pretrained_backbone", None)